SPU launches a wet-lab prediction benchmark for frontier AI models
UniPat AI has introduced Science Prediction Universe (SPU), a benchmark with 116 tasks and 4,207 wet-lab experiments spanning chemistry, materials science, biology and medicine. According to QbitAI’s report, models must independently research experimental conditions and predict measured outcomes; in the first evaluation, Claude Opus 5.5 led with an average score of 0.486, while raising GPT-6 Astra’s inference effort left its prediction performance largely unchanged.
SPU gives research-model developers a standardized way to measure whether AI can predict real experimental outcomes, and its initial results show that more inference effort and compute do not automatically produce better predictions.
「史上最严」湿实验预测基准发布!推理拉满、算力翻5倍也白搭 – 量子位
< img id="wx_img" src="https://www.qbitai.com/wp-content/uploads/imgs/qbitai-logo-1.png" width="400" height="400">
「史上最严」湿实验预测基准发布!推理拉满、算力翻5倍也白搭

斯坦福、哈佛、康奈尔……十余家顶尖团队齐坐一桌了
#UniPat AI#Science Prediction Universe (SPU)#Claude Opus 5.5#GPT-6 Astra#Qwen3.8#Stanford University#Cornell University#Harvard University#FutureHouse